I had the pleasure of attending the Microsoft AI Days event at the Georgia International Conference Center. The event is designed for business leaders, partners, practitioners, and the developer community to come together to learn about the latest technology and innovations from Microsoft. The event featured customer success stories, lessons learned and expert-led technical workshops.
The Question has Become How
The keynote session featured Microsoft executives Ben Grimes, Corporate VP US Manufacturing and Jared Spataro, CMO, AI at Work. Over the past 18 months, customer questions have evolved from “What is AI?” to “How can we implement and benefit from AI?” Three customer representatives from Eaton, Stanley Black and Decker, and Coca Cola shared their experiences. Key takeaways included:
- Adopting artificial intelligence is a journey with a steep learning curve.
- The large language model (LLM) is a very small piece of the puzzle.
- The tools and techniques for reliable AI are multifaceted and complex.
Even with all of its challenges, the panelists agreed that the benefits of AI far outweigh the costs and struggles on the journey.
Assistants, Copilots and AI Agents
The evolution of artificial intelligence is rapid and speeding up exponentially. Current AI usage can be broken into three categories:
- Assistants help with research and provide information for users. Examples include AI assisted search in your browsers or common uses of ChatGPT and other LLMs..
- Copilots go further by assimilating information and making recommendations. They may use generative AI to draft a blogpost, grant or other document. They also provide instruction for how to perform tasks.
- Agents go even further by taking action independently based upon conditions, triggers and other mechanisms.
AI Agents are workers for your company. With a well defined process and the proper training AI Agents are driving massive increases in productivity by performing tasks quicker and without human intervention.
Microsoft Copilot Studio for AI Agents
Microsoft Copilot Studio is the low-code platform for building and deploying AI-driven agents that can interact with users via text or voice. Artificial intelligence is complex and multi-layered. Beneath the easy to use interfaces of Teams or Outlook lay the databases, security, data source and intelligence tools necessary to bring an agent to life.
The Copilot Studio interface leads a designer through a series of questions for quick development of simple agents. You can start with a template to rapidly deploy agents to your teams with natural language prompts. Enhance the agent capability by pointing to a website for content. Voila, you are now an AI developer!
If it was only that simple! Many of the examples demonstrated were no more functional than a Google or Bing search. However, the power is there when you have the skills to dig deeper.
The Azure AI Foundry is the platform used by AI developers to design, develop and manage AI Agents. The foundry is a portal to a rich set of AI capabilities and tools including a unified Software Development Kit (SDK) and Application Programing Interfaces (APIs). Inside the portal programmers gain access to AI Models from Microsoft and others to build programs and other AI components which can be used in the Copilot Studio.
Agentic AI Systems
Jeromey Farmer, Head of Data & AI Americas from Avanade gave a great presentation on “Driving Innovation with Agentic AI.” His practical explanations with real life case studies drove home the “How” for Agentic AI. Specifically, Professor Farmer explained the different types of agents utilized in Agentic AI.
- Utility Agents are purpose built worker agents designed to perform basic tasks autonomously.
- Super Agents understand the user intent and goal. These agents mobilize a configuration of utility agents to achieve a goal.
- Orchestrator Agents coordinate the process across multiple super agents and utility agents to achieve the overall outcome.
In describing five real life use cases we learned how to successfully execute AI projects as well as what pitfalls to avoid. Some great takeaways include:
- KISS – It is easy to overcomplicate projects and to bite off too much in the early stages.
- Everything does not need to be AI. Some Robotic Process Automation (RPA) projects are being called AI. That is OK as not everything needs to be non-deterministic.
- RPA is characterized by programming choices as opposed to an autonomous agent making decisions on its own.
- RPA routines that work well can be incorporated into Orchestrations.
- Don’t forget the human experience. Ensure AI enhance the overall workflow as it cannot replace everything.
- Ensure AI interventions are timely, relevant and non-disruptive to the user’s workflow.
- Let users provide feedback early in the process and show how the AI learn and improves from their input.
- Incorporate users in the validation process to garner greater acceptance of the technology.
Developing Responsible Artificial Intelligence
Microsoft is an early mover and is working hard to develop its platform to make AI available to all companies. Accordingly, they are investing in the tools and educational resources to promote proper governance and controls over artificial intelligence. The fears surrounding AI are based upon real risks of bias, misinformation, and the resulting errors and unintended consequences. Microsoft defines Responsible AI as an approach to developing, assessing, and deploying AI systems in a safe, trustworthy, and ethical way.
Microsoft developed a Responsible AI Standard for building AI systems according to six principles:
- fairness,
- reliability and safety,
- privacy and security,
- inclusiveness,
- transparency, and
- accountability.
As a leader in technology globally, Microsoft has the responsibility to provide safe and secure platforms and tools. As a leader of your company, you have the responsibility to provide efficiency with accurate results for your customers, suppliers, employees and other stakeholders.
Reliability and Confidence
The expectations surrounding Agentic AI are astronomical. Expecting 100% reliability on the first generation of AI is like asking a programmer to write perfect code on the first try. In the early days developers had not developed multi-agent concepts. Now additional tools are used to increase reliability such as:
- Mimic human validation concepts and use of tools like Reflexion within AI.
- Comparing results against multiple Large Language Models (LLMs) for consistency before accepting information as valid.
- Advanced AI testing tools are being used to test AI Agents!
Return on AI Investments
The complexity of AI development along with uncertainty in operating costs keeps AI limited to the large enterprise. Microsoft is trying to “democratize” AI with defined pricing plans and user friendly tools. Unfortunately, the value derived in the examples depicted in AI Days AI’s were from the automation of unique processes. To make Agentic AI generate a positive return on investment in the small to mid-sized business market requires the following pre-requisites:
- A clear understanding of your internal processes.
- Pre-built AI Agents that can adapt to your unique processes.
- Management that understand digital transformation concepts and support change initiatives.
The closer your business process is to a “standard” process within your particular business management system, the more cost effective the AI journey will be. Microsoft Dynamics 365 Business Central incorporates components of Copilot. Larger companies are already starting on the Microsoft AI journey because their ERP is built on the same Azure, Power Platform, technology stack. We also see new entrants to the market like Cumulus.AI for SAP Business One building Agentic AI on the Microsoft platform. As these products mature, a faster positive return on AI investments will become available to more companies.
Next Steps on Your AI Journey
Whether you are just starting to learn or are ready to jump in, now is the time to consider how AI will impact your business. Artificial Intelligence is poised to revolutionize industries by reshaping business strategies focused on efficiency, decision-making, and customer engagement. Companies that fail to integrate AI risk losing competitiveness, especially in a time of inflationary pressures.
All the major software publishers are incorporating AI in their products. Business that can adopt AI tools that work seamlessly with their ERP and CRM systems can enhance productivity while maintaining security and governance over their data. Understanding your business processes and opportunities for automation, enhancement or re-engineering is critical to achieving a rapid return on your AI investments.
At I-BN we offer a variety of Automation Services designed to position your company for the future. We understand the complexity and components of Microsoft Pricing models as well as many of the tools for artificial intelligence in this rapidly changing and expanding marketplace. We are product and brand independent and can guide you on your journey regardless of your core ERP product.
By completing an Automation Exploration Request we will provide you an initial consultation at no charge. If we can’t help you we can likely refer you to a member of our Network of professionals who can!